In-situ monitoring of nitrate in industrial wastewater using Fourier transform infrared attenuated total reflectance spectroscopy (FTIR-ATR) coupled with chemometrics methods

Abstract

Quantitative prediction of nitrate contents in different industrial wastewater was carried out using Fourier transform infrared attenuated total reflectance (FTIR-ATR) spectroscopy. The algorithm of Gaussian deconvolution was applied in the spectral range of 1500–1200 cm−1 to eliminate the background interferences on target information of nitrate, and partial least squares regression (PLSR) model and support vector machine (SVR) model were developed for the prediction of nitrate. The results showed that the PLSR model (Rv2 = 0.921, RMSEv = 0.351 mg/L, RPDv = 3.56) and SVR model (Rv2 = 0.856, RMSEv = 0.473 mg/L, RPDv = 3.15) reached excellent prediction accuracy and robustness for electroplating wastewater, and for metallurgical wastewater the SVR model (Rv2 = 0.916, RMSEv = 1.38 mg/L, RPDv = 3.26) showed a better prediction performance. The PLSR and SVR models exhibited poor prediction accuracy of nitrate for pesticide wastewater and dyeing wastewater due to the strongly interference by carbonate. The spectra pretreatment by deconvolution dramatically improved the prediction models. Therefore, combined with deconvolution spectra pretreatment and chemometrics methods, FTIR-ATR could achieve a fast and effective in-situ monitoring of nitrate in industrial wastewater.